Professional Documents
Culture Documents
DOI
10.1016/j.ijheatfluidflow.2021.108800
Publication date
2021
Document Version
Final published version
Published in
International Journal of Heat and Fluid Flow
Citation (APA)
Blishchik, A., van der Lans, M., & Kenjereš, S. (2021). An extensive numerical benchmark of the various
magnetohydrodynamic flows. International Journal of Heat and Fluid Flow, 90, [108800].
https://doi.org/10.1016/j.ijheatfluidflow.2021.108800
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent
of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Takedown policy
Please contact us and provide details if you believe this document breaches copyrights.
We will remove access to the work immediately and investigate your claim.
A R T I C L E I N F O A B S T R A C T
Keywords: There is a continuous need for an updated series of numerical benchmarks dealing with various aspects of the
MHD magnetohydrodynamics (MHD) phenomena (i.e. interactions of the flow of an electrically conducting fluid and
Magnetic field an externally imposed magnetic field). The focus of the present study is numerical magnetohydrodynamics
Lorentz force
(MHD) where we have performed an extensive series of simulations for generic configurations, including: (i) a
OpenFOAM
Numerical benchmark
laminar conjugate MHD flow in a duct with varied electrical conductivity of the walls, (ii) a back-step flow, (iii) a
multiphase cavity flow, (iv) a rising bubble in liquid metal and (v) a turbulent conjugate MHD flow in a duct with
varied electrical conductivity of surrounding walls. All considered benchmark situations are for the one-way
coupled MHD approach, where the induced magnetic field is negligible. The governing equations describing
the one-way coupled MHD phenomena are numerically implemented in the open-source code OpenFOAM. The
novel elements of the numerical algorithm include fully-conservative forms of the discretized Lorentz force in the
momentum equation and divergence-free current density, the conjugate MHD (coupling of the wall/fluid do
mains), the multi-phase MHD, and, finally, the MHD turbulence. The multi-phase phenomena are simulated with
the Volume of Fluid (VOF) approach, whereas the MHD turbulence is simulated with the dynamic Large-Eddy
Simulation (LES) method. For all considered benchmark cases, a very good agreement is obtained with avail
able analytical solutions and other numerical results in the literature. The presented extensive numerical
benchmarks are expected to be potentially useful for developers of the numerical codes used to simulate various
types of the complex MHD phenomena.
1. Introduction aspects of the MHD phenomena. One of the simplest numerical MHD
benchmarks is a fully developed laminar channel, duct, or pipe flow
One of the pre-requisites to be able to deal with advanced physical subjected to a uniform magnetic field of different orientations, for which
transport phenomena involving the magnetohydrodynamics (MHD) in an exact analytical solution exists, Hartmann and Lazarus (1937),
teractions is to have a well-validated and numerically efficient computer Shercliff (1953). The effects of the non-uniform longitudinal magnetic
code. This still poses a quite challenging task due to a lack of advanced field on a laminar flow of electrically conducting fluid in a pipe were
experimental studies that can provide detailed insights into the flow and recently numerically simulated in Feng et al. (2015). The open-source
electromagnetic parameters that can be used to validate computer computer code OpenFOAM was used and good agreement was ob
codes. The essence of the MHD phenomena is usually associated with a tained between simulations and experiments. The MHD flow in a duct
flow of highly electrically conducting liquid metals, which are, due to with very thin electrically conducting walls was presented in Tao and Ni
their non-transparency, notoriously difficult to study with standard (2013). Instead of fully resolving the wall region, a special type of
laser-based optics diagnostics tools. boundary conditions was applied at the wall/fluid interface that takes
To validate MHD numerical models, we have to rely on analytical into account a finite wall conductivity, as proposed in Walker (1981). It
solutions that are based on significant simplifications. In the present should be noted that this approach can be applied only for a very thin
manuscript, we are revisiting and proposing an extensive list of possible wall thickness and small conductance ratios.
benchmark cases available in the open literature dealing with various Fusion engineering and technology-related research include
* Corresponding author.
E-mail address: S.Kenjeres@tudelft.nl (S. Kenjereš).
https://doi.org/10.1016/j.ijheatfluidflow.2021.108800
Received 16 August 2020; Received in revised form 1 February 2021; Accepted 22 February 2021
Available online 18 May 2021
0142-727X/© 2021 The Author(s). Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
numerous topics dealing with the MHD phenomena. Smolentsev et al. (constant) magnetic field through the Lorentz force. Conservation of
(2015) provided an extensive review of MHD codes for fusion applica mass and momentum are used to describe the MHD flow (under the
tions and selected benchmark problems of importance for fusion appli assumption that the imposed magnetic field is known), and are written
cations. The proposed benchmarks covered a series of 2D and 3D steady as:
and developing MHD flows in both laminar and turbulent regimes, and
∇⋅U = 0 (1)
the final case also included effects of thermal buoyancy. Gajbhiye et al.
(2018) validated their general-purpose solver by analyzing the free ∂U 1 1
convection in a cubical enclosure under a uniform magnetic field and the + (U⋅∇)U = − ∇p + ν∇2 U + (J × B) (2)
∂t ρ ρ
electro-magnetically driven flow in a toroidal duct. The commercial
ANSYS-CFX finite-volume based code was used to simulate a water- where U is velocity, p is pressure, ν is the kinematic viscosity, ρ is den
cooled lithium lead (WCLL) breeding blanket module subjected to a sity, J is the current density and B is the imposed magnetic field. In the
strong uniform magnetic field, Tassone et al. (2017). The commercial momentum equation, the MHD interactions are accounted for through
multi-physics finite-element code COMSOL was successfully applied to the Lorentz force term. In addition to the velocity and pressure, also the
simulate transient natural convection phenomena under influence of the current density (J) needs to be calculated. For the one-way coupled
imposed uniform magnetic field, Sahu and Bhattacharyay (2018). MHD phenomena, i.e. when the following conditions are valid
Validation of the multi-phase MHD flows is a challenging topic. The
number of validation studies dealing with multi-phase MHD phenomena UL ν
Rem = ≪1 and Prm = ≪1 (3)
is significantly smaller compared to single-phase MHD phenomena. The λ λ
analytical solutions for the multi-phase MHD situations are very scarce.
where Rem is the magnetic Reynolds number, Prm is the magnetic Prandtl
One of the recently proposed analytical solutions for a 2D multi-phase
number, L is the characteristic length and λ is the magnetic diffusion, ν is
MHD flow is presented in Righolt et al. (2016), where the elevation of
the kinematic viscosity, the Ohm’s law for a moving conducting fluid is
the liquid–metal/air interface due to the presence of an imposed mag
used
netic field is analytically solved. Numerical simulations of a rising
bubble in the liquid metal subjected to an external homogeneous mag J = σ( − ∇ϕ + U × B) (4)
netic field of different strengths were studied in Shibasaki et al. (2010).
The finite-difference code was used and the terminal bubble velocity where σ is the electrical conductivity of the fluid. By imposing the
dependency on the strength of the imposed magnetic field was analyzed. divergence-free current density condition in the Ohm’s law, i.e.
Finally, the turbulent MHD phenomena require a special solving ∇⋅J = 0 (5)
strategy due to the necessity to properly capture both – the flow and
electromagnetic instabilities. The presence of the fluctuating Lorentz the final Poisson’s equation for the electric potential (ϕ) is obtained and
force requires a proper adaptation of the RANS-type of turbulence can be written as
models (Kenjereš and Hanjalić, 2000; Kenjereš et al., 2004) or applica ( )
tions of the eddy-resolving simulation techniques such as Direct Nu ∇2 ϕ = ∇⋅ U × B (6)
merical Simulations (DNS) or Large Eddy Simulations (LES), Kenjereš In addition to Rem and Prm (given in Eqn. (3)), the hydrodynamic
(2018). Krasnov et al. (2008) compared different sub-grid scale models Reynolds and Hartmann number are used as typical MHD non-
for the MHD LES channel flow and demonstrated ability of the dynamic dimensional parameters:
Smagorinsky model to properly predict the influence of the imposed √̅̅̅̅̅
magnetic field. Chaudhary et al. (2010) used DNS and analyzed how the Re =
UL
, Ha = BL
σ
(7)
increasing strength of a transverse magnetic field could influence the ν ρν
turbulence in the square duct flow. Mao et al. (2017) simulated the MHD
flow in the insulated squared duct with different sub-grid scale models 2.2. Governing equations for a multi-phase MHD: volume of fluid method
and compared data with the DNS results from the previous research of
Chaudhary et al. (2010). Additionally, Mao et al. (2017) varied the In the current study, the Volume of Fluid (VOF) method is applied to
Hartman number, showing how the turbulence is being suppressed by the multi-phase MHD flow simulations. In addition to the Lorentz force,
the imposed magnetic field. also the surface-tension and gravitational forces need to be included into
The main goal of the present study is to obtain and validate results the momentum equation:
from our newly developed OpenFOAM solver over a range of various
∂U 1
magnetohydrodynamic flows, and based on these findings, to propose an + (U⋅∇)U = − ∇p + νav ∇2 U
∂t ρav
extensive numerical MHD benchmark, which can be potentially useful (8)
1 ( )
for developers of the computer codes for simulations of the MHD phe + (J × B) + f g + γk∇α
ρav
nomena. We are primarily focusing on the influence of the finite electric
conductivity of surrounding walls and the multiphase aspects of the
where f g is the gravity force term, γ is the surface tension, k is the cur
MHD phenomena. We have analyzed the following situations: (i) a
vature of the interface (calculated as k = ∇⋅|∇ ∇α
α|), νav is the phase aver
laminar duct flow with finite conductivity of surrounding walls, (ii) a
laminar back-step flow, (iii) a shallow 2D multi-phase cavity, (iv) a aged viscosity (calculated as νav = α⋅ν1 + (1 − α)⋅ν2 , where ’1’ and ’2’
rising bubble in the liquid metal, and, finally, (v) a turbulent duct flow are phase indicators), ρav is the phase averaged density (calculated as
with conducting walls. For all mentioned cases we performed a detailed ρav = α⋅ρ1 + (1 − α)⋅ρ2 ) and the volume fraction α is described by the
comparative assessment against available analytical solutions or/and following transport equation:
numerical results presented in the literature. ∂α
( )
+ ∇⋅ αU + ∇⋅(Ur α(1 − α)) = 0 (9)
∂t
2. Governing equations and numerical details
where Ur is the artificial compression velocity used for the interface
2.1. Governing equations for a single-phase MHD sharpening, which is calculated as:
[ ( )]
|ψ | |ψ |
We consider an incompressible electrically conductive fluid with Ur = nf min Cα ⃒⃒ ⃒⃒, max ⃒⃒ ⃒⃒ (10)
liquid metal properties. The fluid is affected by the imposed external Sf Sf
2
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
where nf is the normal vector of the cell surface, ψ is the mass flux
through the face, Sf is the cell surface area, and Cα is a coefficient that is
used to control the interface thickness. There is no artificial interface
compression when Cα = 0. In order to control the spurious velocities
which appear near the interface due to the sharp change of α, the volume
fraction function is smoothed by the following Laplacian filter (Hoang
et al., 2013; Mukherjee et al., 2018): Fig. 1. Sketch of the fluid/wall interface condition for the conjugate
MHD problem.
∑
n
(αf Sf )
(11)
f =1
α̃c = ∑n ψ mhd = σf (U × B)f ⋅Sf (17)
(Sf )
f =1
where the cell -face electric conductivity (σ f ) is calculated by
where α ̃ is the resulting smooth volume fraction function, while sub applying the harmonic average between different phases, and (Sf ) is
scripts c and f indicate the cell center and cell face, respectively. Using the cell surface area vector.
the smooth function ̃ α in Eqn. (9), instead of the original function α will 2. Use Eq. (17) to solve the discretized electrical potential equation and
suppress these parasitic velocities. In the current study, the filter (11) is find electric potential (ϕ) at the cell centers:
applied twice for each time step. ∑
m ∑
m
σf ∇nf ϕ⋅|Sf | = ψ mhd (18)
f =1 f =1
2.3. The eddy-resolving MHD turbulence: large Eddy simulation
where ‘m’ indicates the number of cell faces.
Turbulence modeling is performed by the Large Eddy Simulation 3. Calculate the current density flux at cell faces using the surface-
(LES) method which is a good alternative to a more computationally normal gradient of electric potential (ϕ):
expensive Direct Numerical Simulations (DNS). We apply the spatial ⃒ ⃒
Jn = − σ f ∇nf ϕ⋅⃒Sf ⃒ + ψ mhd (19)
filtering operation to Eqn. (2), which finally can be written as:
∂U 1 1 where (Jn ) is the cell face normal component of the current density.
+ (U⋅∇)U = − ∇p* + ν∇2 U − ∇⋅τsgs + (J × B) (12) 4. Finally, use the current density flux from Eq. (19) and calculate the
∂t ρ ρ
fully conservative form of the Lorentz force as:
where (’’) indicates the spatially filtered value and (τsgs ) is the sub-grid
1 ∑ m
where (Ωc ) is the volume of cell, (rc ) is the cell center distance vector
where S is the modulus of the strain rate tensor, νsgs is the subgrid scale
and (rf ) is the face center distance vector.
turbulent viscosity, and Cs is Smagorinsky coefficient. In the present
work, we have adopted the dynamic approach to locally estimate values
2.4.2. Conjugate MHD: taking into account electric conductivity and
of the Smagorinsky coefficient, Lilly (1992), as follows:
thickness of surrounding walls
〈 〉
1 Lij Mij The finite electric conductivity and finite thickness of surrounding
C2S = 〈 〉 (14) walls have a significant impact on the fluid flow. This is due to the effects
2 Mij Mij
of the current density transfer between a liquid layer and solid walls,
which is directly influencing the intensity and direction of the local
Lij = Ũ ̃ ̃
i Uj + Ui Uj (15)
Lorentz force in the near-wall region. To include the fluid/wall interface
effects, we have developed an approach similar to traditional conjugate
Mij = Δ2 S̃ ̃2 ̃
Sij + Δ S̃Sij (16) heat transfer, but now instead of the heat flux transfer, we focus on the
distribution of the electric potential and current density in both do
where Δ is the first filter (calculated as Δ = (Δx Δy Δz )1/3 ), Δ
̃ is the mains. Transport equations of the electric potential in liquid (L) and
second filter (calculated as Δ = 2Δ) and ’〈…〉’ means the local spatial
̃ solid (S) wall domains can be written as:
averaging over the cell faces.
∇⋅(σL ∇ϕL ) = ∇⋅(σL (U × B)) (21)
3
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 3. The contours of the streamwise velocity (top row) and electric potential with current density streamlines (bottom row) in the center of the conjugate MHD
duct flow: (a), (e) Ha = 0. (b), (f) Ha = 100, fully insulated walls (Cd = 0). (c), (g) Ha = 100, arbitrary conductive walls (Cd = 0.1). (d), (h) Ha = 100, fully
conductive walls (Cd →∞).
4
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 4. The streamwise velocity profiles along y-axis (between Hartmann walls) and z-axis (between Shercliff walls) in the duct at various Cd and Ha.
Fig. 6. The streamwise velocity profiles along z-axis (between Shercliff walls) in the duct with arbitrary conductivity walls, (Cd = 0.05) for Ha = 5 × 103 (a) and
Ha = 104 (b), respectively.
5
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
{
12(y − 1)(1 − 2y), L/2 < y < L
u(x = 0, y) = (27)
0, 0 < y < L/2
Fig. 8. The contours of the non-dimensional horizontal (streamwise) velocity (u/u0 ) for Re = 800 and different Ha. (a) Ha = 0, (b) Ha = 5, (c) Ha = 10, (d) Ha =
50.
6
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 10. The vertical profiles of the non-dimensional horizontal (u/u0 ) and vertical (v/u0 ) velocity components at various Ha and two values of Re: Re = 300 (a–b)
and Re = 800 (c–d). Comparison between the reference study based on the Local Radial Basis Function Collocation Method (LRBFCM) (Mramor et al., 2014) and the
present Finite Volume Method (FVM) results.
Fig. 11. The non-dimensional horizontal (u/u0 ) and vertical (v/u0 ) velocity profiles at the exit plane for various meshes at Re = 800 and Ha = 5 compared to the
reference solution (LRBFCM, Mramor et al., 2014).
7
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 14. The free-surface elevation for various Ca and Bo. Comparison between the present simulations (CFD) and analytical solution of Righolt et al. (2016).
8
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 15. The profiles of the velocity components in the proximity of the side-wall extracted along the y = d/2 line (a), (b), and in the center of the cavity extracted
along the x = 0 line (c): Re* = A, Ha* = 1, Bo = A2 , Ca = A4 .
study the influence of the magnetic field strength on the rising bubble
behavior. The electrical conductivity ratio is σ G /σ L = 2.49⋅10− 7 . The Fig. 17. The sketch of the simulation domain for the rising bubble in a liquid
orthogonal mesh is created with the mesh size (Nx × Ny × Nz = 60 × metal subjected to an external (axial) magnetic field.
180 × 60), identical to the mesh used in Shibasaki et al. (2010). The
second-order linear-upwind scheme is used for the convective terms in opposite directions above and below the bubble. The velocity contours
both momentum and volume fraction equations, whereas the backward portray an updraft region in the center of the domain – above and below
scheme is used for time integration. Because of a sharp jump of the the bubble, whereas the down-drafts are generated along the side walls.
electrical properties at the phase interface, we have applied the har Contours of the pressure exhibit almost linear distribution in the vertical
monic interpolation scheme for the electric conductivity. For this case, direction, with small deviations in the proximity of the bubble surface. It
the interface compression coefficient (Cα ) had stronger effect on the final can be seen that the resulting shape of the bubble strongly depends on
shape of the rising bubble. The selected value of Cα = 0.1 proved to be a the imposed magnetic field strength, Fig. 19. The higher Ha leads to the
good choice for both multi-phase benchmarks presented here. The ob bubble stretching in the direction of the imposed magnetic field (y-di
tained characteristic bubble shape, current density streamlines, contours rection) and to a reduction of its rising velocity. We compare our results
of the vertical velocity and pressure in the central vertical plane at an with a numerical study of Shibasaki et al. (2010) who applied the finite-
arbitrary time instant t/t * = 0.02 and for Ha = 50, are shown in Fig. 18. difference (FDM) multi-phase MHD code. Comparison of the computed
The current density streamlines form close loops around the bubble with terminal velocity for different values of Ha is shown in Fig. 20. After an
9
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 18. The bubble shape (extracted as the isosurface of the volume fraction α = 0.5) with superimposed streamlines of the total current density (a), (b) contours of
the non-dimensional vertical velocity (uy /u* ) in the central vertical plane, (c) contours of the non-dimensional pressure field (p/p* ) in the central vertical plane - all at
t/t * = 0.02 and for Ha = 50.
Fig. 19. The bubble shape (identified as the isosurface of the volume fraction α = 0.5) and its location at time instant t/t * = 0.025 for various Ha: Ha = 0, 50, 100,
200 (a–d), respectively.
Table 1
The reattachment position (at y/L = 0 for Re = 300 and 800,
and 0⩽Ha⩽100).
Present LRBFCM, Mramor et al. (2014)
Re Ha x/L x/L
10
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Table 2
The non-dimensional terminal velocity at Ha = 50 and Ha = 200
for different meshes. Comparisons with values presented in the
finite-differences based method (FDM) of Shibasaki et al. (2010).
Ha The non-dimensional Present Shibasaki et al.
terminal velocity, (uy /u* ) (2010)
other surfaces are walls with imposed no-slip velocity boundary condi
tions. The lower and upper walls (Shercliff walls) are fully electrically
insulated (∂ϕ/∂n = 0 and Cd = 0). The front and back walls (perpen
dicular to the imposed magnetic field – Hartmann walls) are considered
to have three different types of electric boundary conditions: (i) the
finite conducting walls with the wall conductance parameter Cd =
(σ S dS )/(σL L) = 0.05, (ii) fully electrically insulated walls, and (iii) fully
conductive walls (ϕ = 0 and Cd →∞). We apply the wall-resolving dy
namic large-eddy simulation (LES) approach. The numerical mesh
Fig. 21. Sketch of the simulation domain for the fully-developed (periodic flow )
contains (Nx × Ny × Nz = 240 × 120 × 120 fluid = (3.456 × 106 ) CVs
in the x-direction) turbulent MHD duct flow with Hartmann walls with finite )
thickness (ds ) and electric conductivity (σ S ). and (Nx × Ny × Nz = 240 × 12 × 120 wall = (0.3456 × 106 ) CVs in the
fluid and wall regions, respectively. The non-dimensional mesh pa
initial slight increase in the terminal velocity for intermediate values of rameters are Δy+ wall = Δzwall ≈ 0.6, Δycore = Δzcore ≈ 6 and Δx ≈ 25.
+ + + +
Ha < 50, a gradual decrease is obtained with a further increase of the The central differencing scheme (CDS) is used for spatial discretization
imposed magnetic field. The agreement between the current simulations and the second-order backward scheme for temporal discretization. The
and data presented in Shibasaki et al. (2010) is good up to Ha = 50. flow is defined with the following set of the non-dimensional parame
After reaching this peak value, larger differences are observed, but ters: Re = 5602 and Ha = 21.2. Furthermore, a simulation with Ha = 0
qualitatively similar trends are observed. Differences for larger values of is performed in order to provide a comparison with the non-MHD
Ha number can be partially explained by the use of different dis neutral case. The selected value of the Reynolds number assures that a
cretization approaches (the present finite-volume vs. finite-difference of fully developed turbulence is generated and maintained. All simulations
Shibasaki et al. (2010)), the application of different convective schemes are statistically averaged over at least 100 flow-through times, and the
(the present second-order linear-upwind vs. the third-order UTOPIA spatial averaging procedure is applied to accelerate the convergence of
scheme of Shibasaki et al. (2010)), as well as due to the absence of the the flow statistics. The instantaneous coherent structures colored by
mesh-dependency study of Shibasaki et al. (2010)). We also performed streamwise velocity for various Ha, and Hartmann walls conductivities
additional simulations with a second-order quadratic-upwind scheme (expressed through the wall-conductivity parameter, Cd ), are shown in
for convective terms in momentum equations, but this resulted in mar Fig. 22. Under the action of the imposed transverse magnetic field, by
ginal differences of rising velocity (less than 1%) compared to the linear- changing the electric properties of the walls, the coherent structures
upwind scheme (see Table 1). start to be suppressed in the proximity of Hartmann walls, as seen from
Finally, we complete a mesh-dependency study for two different the side-views of the duct shown in Fig. 22(a)–(c).
Hartmann numbers Ha =50 and 200, and three meshes: (i) the coarse Contours of the long-term time-averaged streamwise velocity, tur
mesh (M1) (Nx × Ny × Nz = 30 × 90 × 30) = (0.081 × 106 )total CVs, (ii) bulent kinetic energy, and electric potential, for various wall conduc
the present mesh (M2) (Nx × Ny × Nz = 60 × 180 × 60) = tivities, are shown in Fig. 23. Starting from a symmetrical distribution,
contours of the mean streamwise velocity start to be suppressed in the
(0.64 × 106 )total CVs and (iii) the fine mesh (M3) (Nx × Ny × Nz = 120 ×
direction of the imposed magnetic field (y-direction). This behavior is
360 × 120) = (5.1 × 106 )total CVs. Results in Table 2 demonstrate that
caused by the reorganization of the electric current density streamlines.
the finest mesh (M3) provides the best agreement with the reference
In contrast to the fully closed current loops in the fluid region (for
data. However, the difference in terminal velocity values between in
electrically insulated walls), a finite electric conductivity of walls makes
termediate (M2) and fine (M3) mesh is only 1%, while the total number
that current density loops also enter these regions, causing significant
of CVs is four times larger. Based on this small difference, we conclude
changes in resulting Lorentz force components in the y- and z-directions,
that results are grid independent already at the mesh (M2).
respectively.
The contours of the turbulent kinetic energy portray the reorgani
3.5. A conjugate MHD duct flow in a fully developed turbulent regime
zation from fully symmetrical distributions for the neutral case with
characteristic peaks in the proximity of duct walls, Fig. 23(e), to the non-
The final test case is a conjugate MHD square duct flow in a fully
symmetrical distributions for non-insulated walls, Fig. 23(f)–(h). It can
developed turbulent regime. The duct height and width are L, and its
be seen that the levels of turbulent kinetic energy are suppressed in the
length is 16L. The imposed magnetic field is aligned with the y-axis and
proximity of Hartmann walls for the fully insulated and walls with finite
perpendicular to the flow direction, Fig. 21. The periodic boundary
conductivity, Fig. 23(f)–(g). At the same time, distributions of the
conditions are imposed in the streamwise (x-coordinate) direction. All
11
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Fig. 23. The long-term time-averaged contours of the streamwise velocity (top row), turbulent kinetic energy (middle row) and electric potential (bottom row) in the
central vertical plane of the duct shown in Fig. 21, for a fixed value of Re = 5602: (a–e–i) Ha = 0, (b–f–j) Ha = 21.2 with fully electrically insulated walls (Cd = 0),
(c–g–k) Ha = 21.2 with arbitrary conducting walls (Cd = 0.05), (d–h–l) Ha = 21.2 with fully conducting walls (Cd →∞).
Fig. 24. The long-term time-averaged non-dimensional streamwise velocity (the semi-log plots of U+ vs. y+ and z+ ) profiles (where U+ = U/Uτ ,y+ = yUτ /ν, and) in
the proximity of Hartmann (a) and Shercliff wall (b), respectively.
turbulent kinetic energy along Shercliff walls are just slightly affected by profiles, Figs. 24 and 25. The present results obtained with the dynamic
changing Hartmann walls conductivity, Fig. 23(f)–(g). Interestingly, for LES approach are validated against two reference studies: Gavrilakis
the fully conducting walls, the turbulence kinetic energy along Shercliff (1992) who simulated an MHD neutral turbulent duct flow, and
walls is reduced in comparison to values along Hartmann walls, Fig. 23 Chaudhary et al. (2010) who simulated MHD turbulent duct flow at
(h). Ha = 21.2 with fully electrically insulated walls – both using the fully-
The contours of the electric potential also illustrate significant resolving Direct Numerical Simulations (DNS) approach. The time-
changes between the fully electrically insulated Fig. 23 and fully con averaged mean streamwise velocity profiles in the proximity of Hart
ducting Hartmann walls Fig. 23(l), with the latter exhibiting signifi mann and Shercliff walls are shown in Fig. 24(a) and (b), respectively. It
cantly more pronounced non-uniform distribution in the vertical can be seen that a very good agreement between the present and DNS
direction. Next, we move to a more detailed comparison of the charac results from the literature is obtained at both locations. The profiles of
teristic long-term time-averaged first- and second-order statistics the non-dimensional rms values of the fluctuating streamwise velocity
12
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
√̅̅̅̅̅̅̅
2
Fig. 25. The non-dimensional rms of streamwise velocity ( u′ /Uτ ) in the proximity of Hartmann (along y/L) (a) and Shercliff (along z/L) wall (b), respectively, for
different Hartmann wall conductivities.
along the identical locations reveal an interesting behavior, Fig. 25. In References
the proximity of Shercliff walls, an increase of the wall conductivity
produced a gradual decrease of the rms values, Fig. 25(b). In contrast to Hartmann, J., Lazarus, F., 1937. Theory of the laminar ow of an electrically conducting
liquid in an homogeneous magnetic field. Kongelige Danske Videnskabernes Selskab
this behavior, the distributions in the proximity of Hartmann walls Matematisk 15, 1–28.
indicate an initial suppression for the fully electrically insulated walls, Shercliff, J.A., 1953. Steady motion of conducting fluids in pipes under transverse
followed by an increase for the fully conducting walls, Fig. 25(a). Again, magnetic fields. Math. Proc. Cambridge Philos. Soc. 49, 136–144. https://doi.org/
10.1017/S0305004100028139.
a good agreement with available DNS references (for the non-MHD sit Feng, J., Chen, H., He, Q., Ye, M., 2015. Further validation of liquid metal MHD code for
uation and the MHD case with fully insulated walls) is obtained con unstructured grid based on OpenFOAM. Fusion Eng. Design 100, 260–264. https://
firming suitability of here used dynamic LES approach. doi.org/10.1016/j.fusengdes.2015.06.059.
Tao, Z., Ni, M., 2013. Benchmark solutions for MHD solver development. Sci. China Phys.
Mech. Astron. 56, 378–382. https://doi.org/10.1007/s11433-013-4997-5.
4. Summary and conclusion Walker, J., 1981. Magnetohydrodynamic flow in rectangular ducts with thin conducting
walls. J. Mecanique 20, 79–112.
Smolentsev, S., Badia, S., Bhattacharyay, R., Bühler, L., Chen, L., Huang, Q., Jin, H.-G.,
We have presented a comprehensive numerical benchmark study
Krasnov, D., Lee, D.-W., Valls, E.M.D.L., et al., 2015. An approach to verification and
addressing a range of single- and multi-phase one-way coupled MHD validation of MHD codes for fusion applications. Fusion Eng. Design 100, 65–72.
flows. The single-phase cases included the conjugate MHD flows in ducts https://doi.org/10.1016/j.fusengdes.2014.04.049.
with varied electric conductivity of the wall – in both laminar and tur Gajbhiye, N.L., Throvagunta, P., Eswaran, V., 2018. Validation and verification of a
robust 3-D MHD code. Fusion Eng. Design 128, 7–22. https://doi.org/10.1016/j.
bulent flow regimes, and the laminar back-step flow subjected to a fusengdes.2018.01.017.
transverse magnetic field. The multi-phase cases covered a two- Tassone, A., Caruso, G., Nevo, A.D., Piazza, I.D., 2017. CFD simulation of the
dimensional MHD cavity and a rising bubble in a liquid metal flows – magnetohydrodynamic flow inside the WCLL breeding blanket module. Fusion Eng.
Design 124, 705–709. https://doi.org/10.1016/j.fusengdes.2017.05.098.
both simulated with the volume of fluid (VOF) approach. We have Sahu, S., Bhattacharyay, R., 2018. Validation of COMSOL code for analyzing liquid metal
implemented an extended set of MHD transport equations in the open- magnetohydrodynamic flow. Fusion Eng. Design 127, 151–159. https://doi.org/
source code OpenFOAM. Our particular focus was to extend the exist 10.1016/j.fusengdes.2018.01.009.
Righolt, B., Kenjereš, S., Kalter, R., Tummers, M., Kleijn, C., 2016. Analytical solutions of
ing set of MHD benchmarks and to provide a detailed comparison with one-way coupled magnetohydrodynamic free surface flow. Appl. Math. Model. 40
similar studies in the literature. We also proposed a novel methodology (4), 2577–2592. https://doi.org/10.1016/j.apm.2015.09.101.
and benchmark for a conjugate MHD in a turbulent duct flow with an Shibasaki, Y., Ueno, K., Tagawa, T., 2010. Computation of a rising bubble in an enclosure
filled with liquid metal under vertical magnetic fields. ISIJ Int. 50 (3), 363–370.
arbitrary wall conductivity (expressed in terms of the wall conductance https://doi.org/10.2355/isijinternational.50.363.
parameter). For the multi-phase flows, we have introduced a recently Kenjereš, S., Hanjalić, K., 2000. On the implementation of effects of Lorentz force in
proposed analytical solution of a two-dimensional partially-filled cavity turbulence closure models. Int. J. Heat Fluid Flow 21 (3), 329–337. https://doi.org/
10.1016/s0142-727x(00)00017-5.
flow subjected to an external magnetic field. An excellent agreement
Kenjereš, S., Hanjalić, K., Bal, D., 2004. A direct-numerical-simulation-based second-
was obtained for all cases for which analytical solutions are available. moment closure for turbulent magnetohydrodynamic flows. Phys. Fluids 16 (5),
For considered test cases without analytical solutions, a very good 1229–1241.
agreement was obtained with available numerical studies from the Kenjereš, S., 2018. On modeling and eddy-resolving simulations of flow, turbulence,
mixing and heat transfer of electrically conducting and magnetizing fluids: a review.
literature. It is concluded that here developed and validated version of Int. J. Heat Fluid Flow 73, 270–297. https://doi.org/10.1016/j.
the computer code can be used for advanced fundamental and indus ijheatfluidflow.2018.09.003.
trial/technological studies involving various aspects of the MHD Krasnov, D., Zikanov, O., Schumacher, J., Boeck, T., 2008. Magnetohydrodynamic
turbulence in a channel with spanwise magnetic field. Phys. Fluids 20 (9), 095105.
phenomena. https://doi.org/10.1063/1.2975988.
Chaudhary, R., Vanka, S.P., Thomas, B.G., 2010. Direct numerical simulations of
Declaration of Competing Interest magnetic field effects on turbulent flow in a square duct. Phys. Fluids 22 (7),
075102. https://doi.org/10.1063/1.3456724.
Mao, J., Zhang, K., Liu, K., 2017. Comparative study of different subgrid-scale models for
The authors declare that they have no known competing financial large eddy simulations of magnetohydrodynamic turbulent duct flow in OpenFOAM.
interests or personal relationships that could have appeared to influence Comput. Fluids 152, 195–203. https://doi.org/10.1016/j.compfluid.2017.04.024.
Hoang, D.A., Steijn, V.V., Portela, L.M., Kreutzer, M.T., Kleijn, C.R., 2013. Benchmark
the work reported in this paper. numerical simulations of segmented two-phase flows in microchannels using the
volume of fluid method. Comput. Fluids 86, 28–36. https://doi.org/10.1016/j.
Acknowledgments compfluid.2013.06.024.
Mukherjee, S., Zarghami, A., Haringa, C., van As, K., Kenjeres, S., van den Akker, H.,
2018. Simulating liquid droplets: a quantitative assessment of lattice Boltzmann and
This project has received funding from the European Union’s Hori Volume of Fluid methods. Int. J. Heat Fluid Flow 70, 59–78. https://doi.org/
zon 2020 research and innovation program TOMOCON (Smart Tomo 10.1016/j.ijheatfluidflow.2017.12.001.
Lilly, D.K., 1992. A proposed modification of the Germano subgrid-scale closure method.
graphic Sensors for Advanced Industrial Process Control) under the
Phys. Fluids A Fluid Dyn. 4 (3), 633–635. https://doi.org/10.1063/1.858280.
Marie Sklodowska-Curie grant agreement No. 764902.
13
A. Blishchik et al. International Journal of Heat and Fluid Flow 90 (2021) 108800
Ni, M.-J., Munipalli, R., Huang, P., Morley, N.B., Abdou, M.A., 2007. A current density Sloan, D., Smith, P., 1966. Magnetohydrodynamic flow in a rectangular pipe between
conservative scheme for incompressible mhd flows at a low magnetic Reynolds conducting plates. Z. Angew. Math. Mech. 46, 439–443. https://doi.org/10.1002/
number. Part II: on an arbitrary collocated mesh. J. Comput. Phys. 227 (1), 205–228. zamm.19660460705.
https://doi.org/10.1016/j.jcp.2007.07.023. Mramor, K., Vertnik, R., Šarler, B., 2014. Simulation of laminar backward facing step
Weller, H.G., Tabor, G., Jasak, H., Fureby, C., 1998. A tensorial approach to flow under magnetic field with explicit local radial basis function collocation
computational continuum mechanics using object-oriented techniques. Comput. method. Eng. Anal. Boundary Elem. 49, 37–47. https://doi.org/10.1016/j.
Phys. 12 (6), 620. https://doi.org/10.1063/1.168744. enganabound.2014.04.013.
Issa, R., Gosman, A., Watkins, A., 1986. The computation of compressible and Gavrilakis, S., 1992. Numerical simulation of low-Reynolds-number turbulent flow
incompressible recirculating flows by a non-iterative implicit scheme. J. Comput. through a straight square duct. J. Fluid Mech. 244 (-1), 101. https://doi.org/
Phys. 62 (1), 66–82. https://doi.org/10.1016/0021-9991(86)90100-2. 10.1017/s0022112092002982.
Hunt, J.C.R., 1965. Magnetohydrodynamic flow in rectangular ducts. J. Fluid Mech. 21,
577–590. https://doi.org/10.1017/S0022112065000344.
14